Intelligent detection and water absorption treatment system for internal water accumulation of unmanned aerial vehicle airspeed meter

Abstract

The present application relates to the technical field of unmanned aerial vehicle fault diagnosis and maintenance, in particular to an unmanned aerial vehicle pitot tube internal water accumulation intelligent detection and water absorption treatment system, comprising: a data acquisition module for collecting humidity, airflow pressure difference and temperature data inside the pitot tube; a feature analysis module for fusion analysis of multi-source data to generate humidity state features and airflow anomaly features; a water accumulation diagnosis module for inputting the above features into a pre-trained water accumulation diagnosis model to calculate real-time water accumulation probability, distribution area estimate and volume estimate; and a disposal control module for generating graded disposal instructions according to the quantitative results. The present application realizes early warning by fusing airflow and humidity features, and realizes quantitative and positioning evaluation of water accumulation state based on an intelligent model, thereby supporting intelligent maintenance from accurate diagnosis to graded disposal, and improving the reliability and maintenance efficiency of the system.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) fault diagnosis and maintenance technology, and in particular to an intelligent detection and water absorption system for water accumulation inside the airspeed indicator of a UAV. Background Technology

[0002] Current methods for detecting water accumulation inside the airspeed indicator of drones primarily rely on installing humidity sensors within the cavity and setting fixed humidity thresholds for monitoring. When the sensor reading exceeds the threshold, the system triggers a simple water accumulation alarm. This is a direct and passive monitoring method, and its technical logic is based on the premise that moisture accumulation has led to a significant change in ambient humidity.

[0003] This conventional technical solution has shortcomings. Humidity threshold alarms cannot detect the early stages of water accumulation, such as the initial stage of condensation or when a small amount of water has infiltrated. Humidity changes may not yet reach the alarm threshold, but they have already potentially affected airflow. A single humidity parameter cannot effectively distinguish between condensate and externally infiltrated liquids, nor can it promptly detect minute airflow changes caused by localized blockages. Existing technology can only provide a binary "present" or "absent" judgment, unable to assess the severity of water accumulation, its specific location, or its approximate volume. This directly leads to simplistic and indiscriminate subsequent response measures, failing to take targeted action based on the actual situation of the water accumulation. This may delay treatment or require unnecessary maintenance, impacting flight safety and equipment lifespan.

[0004] There is a need for a technological solution that can detect abnormal moisture levels inside airspeed sensors earlier and more accurately, and can quantitatively assess their condition. The key issues are how to utilize multi-dimensional sensor information, especially changes in aerodynamic parameters, to achieve early, indirect diagnosis of water accumulation; and how to upgrade simple presence alarms to refined condition assessments that include probability, location, and volume information through intelligent algorithms that integrate multi-source features, thereby providing a clear basis for subsequent differentiated handling procedures. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an intelligent detection and absorption system for water accumulation inside the airspeed meter of unmanned aerial vehicles.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent detection and absorption system for water accumulation inside the airspeed indicator of a UAV, comprising:

[0007] The data acquisition module acquires multi-source environmental sensing data from inside the UAV airspeed meter. The multi-source environmental sensing data includes the humidity value of the airspeed tube cavity collected by the humidity sensor, the air pressure difference value collected by the micro-pressure differential sensor, and the internal temperature reading collected by the temperature sensor.

[0008] The feature analysis module performs fusion preprocessing analysis on the multi-source environmental sensing data to generate humidity state characteristics and airflow anomaly characteristics inside the airspeed meter. The humidity state characteristics include average humidity level, humidity gradient change trend and condensation risk index. The airflow anomaly characteristics include pressure difference fluctuation amplitude, airflow stability coefficient and blockage tendency assessment value.

[0009] The water accumulation diagnosis module inputs the humidity state characteristics and airflow anomaly characteristics into the pre-trained water accumulation diagnosis model to calculate the real-time water accumulation probability inside the airspeed indicator, the estimated value of the water accumulation distribution area, and the estimated water accumulation volume.

[0010] The treatment control module generates graded treatment instructions based on the real-time water accumulation probability, water distribution area estimate, and water volume estimate inside the airspeed meter. The graded treatment instructions include instructions to start the active water absorption program, instructions to start the auxiliary drying program, or instructions to trigger a maintenance alarm.

[0011] As a further aspect of the present invention, the step of fusing and preprocessing the multi-source environmental sensing data to generate humidity state characteristics and airflow anomaly characteristics within the airspeed meter includes:

[0012] The cavity humidity value is subjected to time series smoothing filtering to eliminate instantaneous measurement noise and obtain a smoothed cavity humidity sequence;

[0013] Calculate the average value of the smoothed cavity humidity sequence over a preset analysis period, and use it as the average humidity level;

[0014] The smoothed cavity humidity sequence is analyzed for its rate of change and second derivative during the analysis period to determine whether the humidity is continuously rising, falling, or fluctuating. A humidity change curve is then fitted, and the humidity gradient change trend is extracted from it.

[0015] By combining the internal temperature readings with the smoothed cavity humidity sequence, and based on the dew point temperature calculation model, the probability of condensation occurring on the internal surface of the airspeed meter under the current internal temperature and humidity conditions is assessed, and the condensation risk index is calculated.

[0016] Analyze the fluctuation of the airflow pressure difference over time, calculate its standard deviation and peak-to-peak value, and obtain the amplitude of the pressure difference fluctuation.

[0017] The airflow stability coefficient is calculated based on the degree of deviation between the airflow pressure difference value and the standard pressure difference value under the theoretical no-water-accumulation condition, as well as the frequency characteristics of the pressure difference change.

[0018] Based on the variation pattern of the airflow pressure difference value, combined with the fluid dynamics model, it is determined whether there is a local decrease in flow velocity or airflow separation phenomenon in the airspeed tube, and the blockage tendency assessment value characterizing local blockage is generated.

[0019] As a further aspect of the present invention, the method for calculating the condensation risk index includes:

[0020] Based on the internal temperature reading and the smoothed cavity humidity sequence, the dew point temperature under the current condition is calculated using the internationally recognized dew point calculation formula.

[0021] The difference between the internal temperature reading and the dew point temperature is calculated to obtain the superheat.

[0022] Establish a mapping table between superheat and condensation risk index. In the mapping table, the smaller the superheat, the higher the condensation risk index.

[0023] Based on the calculated superheat value, the corresponding condensation risk index is obtained by querying the mapping table.

[0024] Meanwhile, by taking into account the local temperature differences in different parts of the airspeed gauge, the condensation risk index is spatially weighted and corrected to more accurately reflect the risk level in different areas.

[0025] As a further aspect of the present invention, the step of inputting the humidity state characteristics and airflow anomaly characteristics into a pre-trained water accumulation diagnosis model to calculate the real-time water accumulation probability inside the airspeed indicator, the estimated value of the water accumulation distribution area, and the estimated water accumulation volume includes:

[0026] The water accumulation diagnostic model receives the input of the average humidity level, humidity gradient change trend, condensation risk index, pressure difference fluctuation amplitude, airflow stability coefficient, and blockage tendency assessment value.

[0027] The water accumulation diagnosis model, based on its internal multi-layer neural network structure, first performs feature fusion and dimensionality reduction to generate a comprehensive state vector.

[0028] The comprehensive state vector is matched with the state vectors in the historical water accumulation case library for similarity, and the real-time water accumulation probability inside the airspeed meter is calculated using the built-in logistic regression classifier of the model. The probability value represents the possibility that water accumulation exists at present.

[0029] Meanwhile, the water accumulation diagnostic model estimates the location of water accumulation by using an interpolation algorithm based on the spatial correlation between the condensation risk index and the blockage tendency assessment value, combined with the three-dimensional model of the internal structure of the air velocity meter, and outputs the estimated value of the water accumulation distribution area.

[0030] The water accumulation diagnostic model further calculates the estimated water accumulation volume by using the fluid volume function in conjunction with the deviation between the airflow pressure difference value and the theoretical dry pressure difference value, combined with the internal geometric parameters of the airspeed tube.

[0031] As a further aspect of the present invention, the water accumulation diagnostic model, based on the spatial correlation between the condensation risk index and the blockage tendency assessment value, and combined with a three-dimensional model of the internal structure of the air velocity meter, estimates the location of water accumulation using an interpolation algorithm, and outputs an estimated value of the water accumulation distribution area, specifically including:

[0032] Obtain the coordinate information of each position in the three-dimensional model of the internal structure of the airspeed indicator;

[0033] The condensation risk index is mapped to the corresponding sensor installation location in the three-dimensional model of the internal structure of the airspeed meter;

[0034] The blockage tendency assessment value is mapped to the location of the corresponding airflow monitoring point in the three-dimensional model of the internal structure of the airspeed indicator;

[0035] Calculate the spatial correlation weight between the condensation risk index and the blockage tendency assessment value, wherein the correlation weight is determined based on the distance and numerical correlation between the two feature values ​​in three-dimensional space;

[0036] Based on the associated weights, bilinear interpolation calculations are performed on the mesh nodes of the three-dimensional model of the internal structure of the airspeed indicator.

[0037] A water accumulation probability distribution map of each grid node inside the airspeed meter is generated using an interpolation algorithm;

[0038] Extract continuous areas with a water accumulation probability higher than a set threshold from the water accumulation probability distribution map, and use them as the estimated value of the water accumulation distribution area.

[0039] As a further aspect of the present invention, the training and updating method of the pre-trained water accumulation diagnosis model includes:

[0040] In a laboratory or controlled environment, different degrees of internal water accumulation or high humidity were artificially simulated for different types of UAV airspeed meter samples.

[0041] When simulating various states, corresponding multi-source environmental sensor data are collected synchronously, including cavity humidity value, airflow pressure difference value and internal temperature reading, and the corresponding real water accumulation situation is recorded as a label. The real water accumulation situation includes whether there is water accumulation, the approximate area and volume range of water accumulation.

[0042] The collected labeled sensor data is used to conduct supervised training on the initial neural network model. The model parameters are adjusted through the backpropagation algorithm until the error between the model output water accumulation probability, area estimation and volume estimation and the real label meets the accuracy requirements.

[0043] Deploy the trained model to a real drone system;

[0044] In actual operation, when the system executes the aforementioned maintenance alarm command and the maintenance personnel conduct on-site handling and confirm the water accumulation, the maintenance confirmation result is used as a new real label, which together with the sensor data recorded at that time constitutes a new training sample.

[0045] After periodically or accumulating a certain number of new samples, the water accumulation diagnosis model is incrementally learned or fine-tuned to continuously optimize model performance.

[0046] As a further aspect of the present invention, the step of generating a graded treatment instruction based on the real-time water accumulation probability inside the airspeed indicator, the estimated value of the water accumulation distribution area, and the estimated water accumulation volume includes:

[0047] The probability of water accumulation inside the airspeed meter in real time is compared with a first probability threshold and a second probability threshold, wherein the first probability threshold is lower than the second probability threshold.

[0048] When the probability of water accumulation inside the airspeed meter is lower than the first probability threshold, it is determined that there is no water accumulation or only a trace of moisture, a monitoring command is generated, and the system maintains its normal monitoring state.

[0049] When the real-time probability of water accumulation inside the airspeed meter is between the first probability threshold and the second probability threshold, it is determined that there is an initial risk of water accumulation or moderate humidity.

[0050] In this case, if the estimated water volume is less than a preset volume threshold, the command to start the active water absorption program is generated.

[0051] If the estimated volume of water accumulation is greater than or equal to the preset volume threshold, it is determined that there is a significant risk of water accumulation.

[0052] When the probability of water accumulation inside the airspeed meter is higher than the second probability threshold, it is directly determined that there is a high probability of water accumulation or a serious risk, and the maintenance alarm command is generated.

[0053] For cases where an initial risk of water accumulation or moderate humidity is identified and the estimated water volume is less than the volume threshold, a water absorption strategy for a specific area is embedded in the active water absorption program command, based on the estimated water distribution area.

[0054] As a further aspect of the present invention, after generating the instruction to initiate the active water absorption process, the working method further includes executing the active water absorption process:

[0055] The active water absorption program command is analyzed to determine the water absorption strategy for a specific area, and the target water absorption area and the recommended water absorption flow rate are determined.

[0056] Activate the micro electronically controlled valve and the micro negative pressure pump located near the target water absorption area, wherein the micro negative pressure pump is connected to a specific suction port of the internal cavity of the air velocity meter through a pipeline;

[0057] The opening degree of the micro electronically controlled valve is controlled to match the recommended water suction flow rate, and the micro negative pressure pump is started to generate negative pressure to draw the suspected water or high humidity gas inside the air velocity meter to the external water collection container or drying device.

[0058] During the active water absorption process, the humidity value of the cavity and the airflow pressure difference value are continuously and synchronously acquired and updated.

[0059] The effectiveness of the water absorption treatment is evaluated in real time by analyzing the rate of decrease in the humidity value of the cavity and the trend of the airflow pressure difference value recovering to the standard value.

[0060] As a further aspect of the present invention, the execution of the instruction to initiate the auxiliary drying process includes:

[0061] If, after the active water absorption process is executed, the updated humidity value of the cavity is still higher than the preset safe humidity threshold, or the airflow pressure difference value has not fully returned to the normal range, then it is determined that auxiliary drying is required.

[0062] Generate the instruction to start the auxiliary drying program, which includes a suggested drying temperature and drying duration;

[0063] Activate the electrothermal film or miniature electrothermal element built into or near the airspeed gauge cavity wall;

[0064] The electric heating film or micro electric heating element is controlled to heat at the recommended drying temperature to assist in drying the residual moisture inside the airspeed meter.

[0065] At the same time, the miniature negative pressure pump is kept running at low power to continuously discharge the water vapor that has been heated and evaporated;

[0066] During the assisted drying process, the internal temperature readings and the updated humidity values ​​of the cavity are continuously monitored to ensure that the temperature does not exceed the upper limit of the air velocity meter material's tolerance, and the drying parameters are dynamically adjusted according to the decrease in humidity.

[0067] As a further aspect of the present invention, the execution of the trigger maintenance alarm command includes:

[0068] When the aforementioned maintenance alarm command is generated, the system immediately interrupts any ongoing water absorption or drying process;

[0069] Assemble an alarm information message, the alarm information message including at least the real-time probability of water accumulation inside the airspeed meter, the estimated value of the water accumulation distribution area, the estimated amount of water accumulation volume, the last valid humidity status characteristics, and the airflow anomaly characteristics;

[0070] The alarm information message is sent to the ground control station or the handheld terminal of the maintenance personnel via the airborne data link of the UAV or a dedicated maintenance interface;

[0071] Record all relevant data for this alarm event in the system's local storage unit, including timestamps, all sensor data sequences, and diagnostic model output results;

[0072] Switch the system status to security monitoring mode. In this mode, only basic environmental monitoring functions are maintained, and the system awaits external maintenance intervention instructions.

[0073] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0074] A micro-differential pressure sensor is introduced to monitor airflow differential pressure values ​​and analyze them to generate abnormal airflow characteristics such as differential pressure fluctuation amplitude, airflow stability coefficient, and blockage tendency assessment value. This technique transforms changes in the physical state of the airflow channel into quantifiable diagnostic parameters. Water accumulation or condensation inside the airspeed meter alters airflow dynamics; even if humidity does not reach a threshold, the differential pressure signal will exhibit characteristic fluctuations. Through continuous assessment of airflow stability and blockage tendency, the system can identify channel performance degradation caused by trace moisture accumulation, enabling early risk warnings. Simultaneously, the fusion analysis of airflow characteristics and humidity data effectively distinguishes between increased ambient humidity and substantial water accumulation or localized blockage, reducing the possibility of false alarms or missed alarms from a single humidity sensor and improving the reliability of condition assessment.

[0075] This system employs a pre-trained water accumulation diagnostic model, receiving fused multi-dimensional features and outputting real-time water accumulation probability, estimated distribution area, and estimated volume. This method analyzes the complex relationship between features and water accumulation status through an algorithmic model, achieving both quantitative and spatial diagnostic results. The system not only determines whether water accumulation has occurred but also assesses its probability in probabilistic terms and estimates the approximate location and volume of the water. The refined output based on probability, area, and volume provides a clear basis for subsequent response decisions. The response system can select different response levels based on the water accumulation probability, pinpoint key treatment areas based on the estimated distribution area, and match appropriate water absorption or drying intensities with reference to the estimated volume, thus forming a closed loop from precise perception to precise execution, improving maintenance efficiency and the effectiveness of actions. Attached Figure Description

[0076] Figure 1 This is a timing diagram of the intelligent detection and water absorption treatment system for internal water accumulation in the airspeed meter of a UAV as described in this invention.

[0077] Figure 2 A flowchart for calculating the condensation risk index;

[0078] Figure 3A bar chart showing the water accumulation characteristics in different areas inside the airspeed indicator of a drone;

[0079] Figure 4 A bar chart analyzing the alarm triggering conditions of a UAV airspeed meter water accumulation detection system;

[0080] Figure 5 A bar chart comparing multiple parameters of the active water absorption strategy for UAV airspeed meters. Detailed Implementation

[0081] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0082] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0083] See Figure 1 The data acquisition module reads in real time the humidity value of the airspeed indicator cavity from the built-in humidity sensor, the airflow differential pressure value from the micro-differential pressure sensor, and the internal temperature reading from the temperature sensor. These data together constitute a multi-source environmental sensor data stream. The feature analysis module performs fusion preprocessing analysis on the acquired multi-source environmental sensor data stream. This process includes filtering, calculation, and model evaluation of the raw data, ultimately generating humidity state characteristics and airflow anomaly characteristics that quantitatively describe the internal environmental state. These characteristics are input into the pre-trained water accumulation diagnosis model in the water accumulation diagnosis module. This model calculates and outputs the real-time water accumulation probability inside the airspeed indicator, the estimated value of the water accumulation distribution area, and the estimated water accumulation volume through an internal algorithm. The treatment and control module receives the above diagnostic results, compares and judges them according to a set of preset logical rules, and generates graded treatment instructions with different priorities and operation contents. These instructions are used to control subsequent active water absorption procedures, auxiliary drying procedures, or trigger system alarms.

[0084] In one embodiment of the present invention, see [reference] Figure 2The feature analysis module performs fusion and preprocessing analysis on multi-source environmental sensor data to generate humidity state characteristics and airflow anomaly characteristics inside the airspeed meter. Time-series smoothing filtering is applied to the cavity humidity values ​​to eliminate instantaneous measurement noise and obtain a smoothed cavity humidity sequence. The average value of the smoothed cavity humidity sequence over a preset analysis period is calculated, and this average value is taken as the average humidity level. The rate of change and second derivative of the smoothed cavity humidity sequence over the analysis period are analyzed to determine whether the humidity is continuously rising, falling, or fluctuating, and the humidity gradient trend is extracted through curve fitting. Combining the internal temperature readings with the smoothed cavity humidity sequence, the probability of condensation on the internal surface of the airspeed meter under the current internal temperature and humidity conditions is assessed using a dew point temperature calculation model. This assessment result is quantified as a condensation risk index. The calculation method for the condensation risk index specifically includes: calculating the dew point temperature under the current condition using the internationally recognized dew point calculation formula based on the internal temperature reading and the smoothed cavity humidity sequence; calculating the superheat by the difference between the internal temperature reading and the dew point temperature; establishing a mapping relationship table between superheat and the condensation risk index; the smaller the superheat in the table, the higher the condensation risk index; obtaining the corresponding condensation risk index by looking up the mapping relationship table based on the calculated superheat value; and spatially weighting the condensation risk index by considering the local temperature differences in different parts of the airspeed meter. The fluctuation of the airflow pressure difference over time is analyzed, and its standard deviation and peak-to-peak value are calculated to obtain the pressure difference fluctuation amplitude. The airflow stability coefficient is calculated based on the deviation of the airflow pressure difference from the standard pressure difference under the theoretical no-water-accumulation condition and the frequency characteristics of the pressure difference change. Based on the variation pattern of the airflow pressure difference, combined with the fluid dynamics model, it is determined whether there is a local velocity decrease or airflow separation phenomenon in the airspeed tube, and a blockage tendency assessment value characterizing local blockage is generated.

[0085] In specific implementation, the feature analysis module performs fusion preprocessing analysis on multi-source environmental sensor data to generate humidity state characteristics and airflow anomaly characteristics inside the airspeed meter. The feature analysis module performs time-series smoothing filtering on the cavity humidity values ​​to eliminate instantaneous measurement noise and obtain a smoothed cavity humidity sequence. The feature analysis module calculates the average value of the smoothed cavity humidity sequence within a preset analysis period as the average humidity level. The feature analysis module analyzes the rate of change and second derivative of the smoothed cavity humidity sequence within the analysis period to determine whether the humidity is continuously rising, falling, or fluctuating, and fits a humidity change curve to extract the humidity gradient trend. In specific implementation, the feature analysis module combines the internal temperature reading and the smoothed cavity humidity sequence to assess the possibility of condensation on the internal surface of the airspeed meter under the current internal temperature and humidity conditions using a dew point temperature calculation model, and calculates a condensation risk index. The calculation method for the condensation risk index includes calculating the dew point temperature under the current state using an internationally recognized dew point calculation formula based on the internal temperature reading and the smoothed cavity humidity sequence. In some embodiments, the superheat formula is:

[0086]

[0087] in: Indicates overheating. This indicates the internal temperature reading. This indicates the dew point temperature. Based on internal temperature readings The relative humidity values ​​in the smoothed cavity humidity sequence are determined using the internationally recognized dew point calculation formula. The feature analysis module establishes a mapping table between superheat and condensation risk index, where a lower superheat indicates a higher condensation risk index. The feature analysis module retrieves the corresponding condensation risk index from the mapping table based on the calculated superheat value. It can be understood that the feature analysis module also incorporates local temperature differences in different parts of the airspeed sensor to spatially weight and correct the condensation risk index, thus more accurately reflecting the risk level in different areas.

[0088] In practical implementation, the feature analysis module analyzes the fluctuation of the airflow pressure difference over time, calculates its standard deviation and peak-to-peak value to obtain the pressure difference fluctuation amplitude, and calculates the airflow stability coefficient based on the deviation of the airflow pressure difference from the standard pressure difference under theoretical no-water-accumulation conditions and the frequency characteristics of pressure difference changes. It can be understood that the calculation of the airflow stability coefficient involves frequency domain analysis of the airflow pressure difference value sequence to extract the frequency characteristics of pressure difference changes. Optionally, the feature analysis module, based on the variation pattern of the airflow pressure difference value and combined with a fluid dynamics model, determines whether there is a local velocity decrease or airflow separation phenomenon within the airspeed tube, and generates a blockage tendency assessment value characterizing local blockage.

[0089] In one embodiment of the present invention, the water accumulation diagnosis module inputs humidity state characteristics and airflow anomaly characteristics into a pre-trained water accumulation diagnosis model to calculate the real-time water accumulation probability inside the airspeed indicator, the estimated water accumulation distribution area, and the estimated water accumulation volume. The water accumulation diagnosis model receives input average humidity level, humidity gradient trend, condensation risk index, pressure difference fluctuation amplitude, airflow stability coefficient, and blockage tendency assessment value. Based on its internal multi-layer neural network structure, the model first performs feature fusion and dimensionality reduction to generate a comprehensive state vector. The comprehensive state vector is then matched with state vectors in a historical water accumulation case library for similarity, and the model's built-in logistic regression classifier is used to calculate the real-time water accumulation probability inside the airspeed indicator. Simultaneously, based on the spatial correlation between the condensation risk index and the blockage tendency assessment value, and combined with a three-dimensional model of the airspeed indicator's internal structure, the water accumulation diagnosis model uses an interpolation algorithm to estimate the location of water accumulation and outputs an estimated water accumulation distribution area. The process specifically includes: acquiring the coordinate information of each location in the 3D model of the airspeed indicator's internal structure; mapping the condensation risk index to the corresponding sensor installation location in the 3D model of the airspeed indicator's internal structure; mapping the blockage tendency assessment value to the corresponding airflow monitoring point location in the 3D model of the airspeed indicator's internal structure; calculating the spatial correlation weight between the condensation risk index and the blockage tendency assessment value, which is determined based on the distance and numerical correlation between the two feature values ​​in 3D space; performing bilinear interpolation calculations on the grid nodes of the 3D model of the airspeed indicator's internal structure according to the correlation weights; generating a water accumulation probability distribution map for each grid node inside the airspeed indicator using an interpolation algorithm; and extracting continuous areas with a water accumulation probability higher than a set threshold from the water accumulation probability distribution map as estimated water accumulation distribution areas. The water accumulation diagnostic model further calculates the estimated water accumulation volume based on the deviation between the airflow pressure difference value and the theoretical dry pressure difference value, combined with the internal geometric parameters of the airspeed tube and using the fluid volume function for inversion calculation.

[0090] In practical implementation, the water accumulation diagnosis module inputs humidity status characteristics and airflow anomaly characteristics into a pre-trained water accumulation diagnosis model and calculates the real-time water accumulation probability inside the airspeed indicator, the estimated water accumulation distribution area, and the estimated water accumulation volume. The water accumulation diagnosis model receives input average humidity level, humidity gradient change trend, condensation risk index, pressure difference fluctuation amplitude, airflow stability coefficient, and blockage tendency assessment value. Based on its internal multi-layer neural network structure, the water accumulation diagnosis model first performs feature fusion and dimensionality reduction processing to generate a comprehensive state vector. It can be understood that the multi-layer neural network structure includes fully connected layers and activation functions to achieve feature fusion and nonlinear transformation. Dimensionality reduction is completed through pooling layers or dedicated dimensionality reduction layers in the neural network to generate a lower-dimensional comprehensive state vector. The water accumulation diagnosis model performs similarity matching between the comprehensive state vector and the state vectors in the historical water accumulation case library and uses the model's built-in logistic regression classifier to calculate the real-time water accumulation probability inside the airspeed indicator. The historical water accumulation case library stores historical comprehensive state vectors and their corresponding labels that mark the actual water accumulation state under different operating conditions. In some embodiments, similarity matching is achieved by calculating cosine similarity or Euclidean distance. The logistic regression classifier receives the input features composed of similarity matching results and outputs a real-time water accumulation probability value inside the airspeed indicator that is between zero and one.

[0091] In practical implementation, the water accumulation diagnostic model estimates the location of water accumulation and outputs an estimated water accumulation area based on the spatial correlation between the condensation risk index and the blockage tendency assessment value, combined with the three-dimensional model of the airspeed indicator's internal structure, using an interpolation algorithm. The model obtains the coordinate information of each location within the three-dimensional model of the airspeed indicator's internal structure. In some embodiments, the model maps the condensation risk index to the corresponding humidity sensor installation location in the three-dimensional model of the airspeed indicator's internal structure, and maps the blockage tendency assessment value to the corresponding airflow monitoring point location in the same model. The model calculates the spatial correlation weight between the condensation risk index and the blockage tendency assessment value, determined based on the distance and numerical correlation between the two feature values ​​in three-dimensional space. Based on the correlation weight, the model performs bilinear interpolation calculations on the grid nodes of the three-dimensional model of the airspeed indicator's internal structure. Using an interpolation algorithm, the model generates a water accumulation probability distribution map for each grid node within the airspeed indicator. From this map, the model extracts continuous areas with a water accumulation probability higher than a set threshold as estimated water accumulation distribution areas.

[0092] In one embodiment of the present invention, under laboratory or controlled environmental conditions, different degrees of internal water accumulation or high humidity are artificially simulated for different types of UAV airspeed sensor samples. During the simulation of various states, corresponding multi-source environmental sensor data are simultaneously collected, including cavity humidity values, airflow pressure difference values, and internal temperature readings. The corresponding actual water accumulation conditions are recorded as labels, including whether water accumulation exists, the approximate area of ​​water accumulation, and its volume range. The collected labeled sensor data is used to perform supervised training on an initial neural network model. Model parameters are adjusted using a backpropagation algorithm until the model's output water accumulation probability, area estimation, and volume estimation errors meet the accuracy requirements of the actual labels. The trained model is then deployed to an actual UAV system. During actual operation, when the system executes a maintenance alarm command and maintenance personnel conduct on-site handling and confirm the water accumulation situation, the maintenance confirmation result is used as a new actual label, which, together with the sensor data recorded at that time, constitutes a new training sample. Periodically or after accumulating a certain number of new samples, the water accumulation diagnosis model is incrementally learned or fine-tuned.

[0093] In practice, the training and updating of the pre-trained water accumulation diagnostic model are performed under laboratory or controlled environmental conditions. Different degrees of internal water accumulation or high humidity are artificially simulated for different types of UAV airspeedometer samples to construct training samples. During the simulation, corresponding multi-source environmental sensor data are simultaneously collected, including cavity humidity values, airflow pressure difference values, and internal temperature readings. The corresponding actual water accumulation conditions are recorded as labels, including whether water accumulation exists, the approximate area of ​​water accumulation, and its volume range. It can be understood that the artificially simulated internal water accumulation or high humidity conditions encompass a variety of scenarios, ranging from trace amounts of moisture to localized water accumulation. The recording of actual water accumulation conditions is accomplished through precision measuring instruments or precise control and observation of the simulation process.

[0094] In practice, the collected labeled sensor data is used to perform supervised training on the initial neural network model. The model parameters are adjusted using the backpropagation algorithm until the errors in the model's output water accumulation probability, region estimation, and volume estimation compared to the true labels meet the accuracy requirements. In some embodiments, the backpropagation algorithm calculates the gradient based on the loss function and updates the weights and bias parameters of the neural network model. The overall error loss function used to measure the difference between the model output and the true labels is:

[0095]

[0096] in: This represents the total loss value. This represents the binary cross-entropy loss component for the probability of water accumulation. This represents the regional overlap loss component for the estimated values ​​of the waterlogged distribution area. This represents the mean square error loss component for estimating the volume of accumulated water. , , These are hyperparameters used to balance the weights of different loss components. It can be understood that the accuracy requirement is preset according to the actual application needs, and the training process terminates when the loss function value falls below a preset threshold or the performance metrics on the validation set reach a predetermined standard.

[0097] In practical implementation, the trained model is deployed to the actual unmanned aerial vehicle (UAV) system. During actual operation, when the system executes a maintenance alarm command and maintenance personnel conduct on-site handling and confirm the water accumulation situation, the maintenance confirmation result is used as a new real label, which, together with the sensor data recorded at that time, constitutes a new training sample. In some embodiments, the system automatically records all multi-source environmental sensor data sequences from the moment before the alarm is triggered to the moment of maintenance intervention, as well as the intermediate results output by the diagnostic model. Maintenance personnel input the on-site confirmed water accumulation situation through a dedicated maintenance interface as the new real label. Optionally, the system periodically or after accumulating a certain number of new samples performs incremental learning or fine-tuning on the water accumulation diagnostic model to continuously optimize model performance. The incremental learning process uses online learning algorithms or retrains some parameters of the deployed model based on a new sample set.

[0098] See Figure 3 This is a bar chart showing the water accumulation characteristics in different areas inside the airspeed indicator of a drone. The bottom of the cavity has the highest risk of water accumulation, with a probability of approximately 0.95 and a water volume of approximately 5.5 ml, the highest among all areas. Specifically, the highest value among all areas refers to both the highest "probability of water accumulation" and the highest "estimated water volume." This is related to the physical structure of the bottom of the cavity; its lower position makes it prone to liquid water accumulation due to gravity. The airspeed tube inlet and sensor mounting location have the next highest risk of water accumulation, with a water volume of approximately 4.2 ml at the airspeed tube inlet and approximately 3.8 ml at the sensor mounting location. These two areas are directly exposed to external airflow and are the main locations for water vapor entry and condensation. The drain outlet has the lowest risk of water accumulation, with a probability of approximately 0.45 and a volume of approximately 1.8 ml, indicating that the drainage design in this area can effectively reduce the risk of water accumulation under normal circumstances. Overall, areas with a high probability of water accumulation also have a larger water volume, showing a positive correlation, which is consistent with the theoretical model of the drone airspeed indicator water accumulation detection system.

[0099] In one embodiment of the present invention, the disposal control module generates graded disposal instructions based on the real-time probability of water accumulation inside the airspeed indicator, the estimated value of the water accumulation distribution area, and the estimated water volume. The real-time probability of water accumulation inside the airspeed indicator is compared with a first probability threshold and a second probability threshold, wherein the first probability threshold is lower than the second probability threshold. When the real-time probability of water accumulation inside the airspeed indicator is lower than the first probability threshold, it is determined that there is no water accumulation or only a trace amount of moisture, a monitoring instruction is generated, and the system maintains its normal monitoring state. When the real-time probability of water accumulation inside the airspeed indicator is between the first and second probability thresholds, it is determined that there is an initial risk of water accumulation or moderate moisture. In this case, the estimated water volume is further considered. If the estimated water volume is less than a preset volume threshold, an instruction to initiate an active water absorption procedure is generated; if the estimated water volume is greater than or equal to the preset volume threshold, it is determined that there is a significant risk of water accumulation. When the real-time probability of water accumulation inside the airspeed indicator is higher than the second probability threshold, it is directly determined that there is a high probability of water accumulation or a serious risk, and a maintenance alarm instruction is generated. For situations where initial water accumulation or moderate humidity risk is identified and the estimated water volume is less than the volume threshold, a water absorption strategy targeting a specific area is embedded in the active water absorption program command, based on the estimated water distribution area. After generating the active water absorption program command, the system executes the active water absorption process: it parses the water absorption strategy embedded in the command to determine the target absorption area and recommended absorption flow rate; it activates a miniature electronically controlled valve and a miniature negative pressure pump located near the target absorption area, with the miniature negative pressure pump connected to a specific suction port inside the air velocity meter via a pipeline; it controls the opening of the miniature electronically controlled valve to match the recommended absorption flow rate, while simultaneously activating the miniature negative pressure pump to generate negative pressure, drawing suspected water or high-humidity gas from inside the air velocity meter to an external water collection container or drying device; during the active water absorption process, it continuously acquires updated chamber humidity and airflow pressure difference values; and it analyzes the rate of decrease in chamber humidity and the trend of the airflow pressure difference value recovering to the standard value.

[0100] In specific implementation, the handling control module generates graded handling instructions based on the real-time water accumulation probability inside the airspeed indicator, the estimated water distribution area, and the estimated water volume. The module compares the real-time water accumulation probability inside the airspeed indicator with a first probability threshold and a second probability threshold, where the first probability threshold is lower than the second probability threshold. When the real-time water accumulation probability inside the airspeed indicator is lower than the first probability threshold, the handling control module determines there is no water accumulation or only a trace of moisture and generates a monitoring instruction while maintaining the system's normal monitoring status. When the real-time water accumulation probability inside the airspeed indicator is between the first and second probability thresholds, the handling control module determines there is an initial risk of water accumulation or moderate moisture. In this case, the handling control module further combines the estimated water volume for judgment. In some embodiments, if the estimated water volume is less than a preset volume threshold, the handling control module generates an instruction to initiate an active water suction program; if the estimated water volume is greater than or equal to the preset volume threshold, the handling control module determines there is a significant risk of water accumulation. When the real-time water accumulation probability inside the airspeed indicator is higher than the second probability threshold, the handling control module directly determines there is a high probability of water accumulation or a serious risk and generates a maintenance alarm instruction. For situations where initial water accumulation or moderate humidity risk is identified and the estimated water volume is less than the volume threshold, the treatment control module incorporates a water absorption strategy for a specific area into the active water absorption program initiation command, based on the estimated water distribution area. See Table 1 for the generation logic of the tiered treatment command.

[0101] Table 1: Logical Comparison Table of Tiered Handling Instructions

[0102]

[0103] In practical implementation, after generating the instruction to initiate the active water suction program, the system executes the active water suction process. The system parses the water suction strategy embedded in the active water suction program instruction for a specific area and determines the target water suction area and the recommended water suction flow rate. The system activates a miniature electronically controlled valve and a miniature negative pressure pump located near the target water suction area. The miniature negative pressure pump is connected to a specific suction port inside the air velocity meter's internal cavity via a pipeline. The system controls the opening degree of the miniature electronically controlled valve to match the recommended water suction flow rate while simultaneously activating the miniature negative pressure pump to generate negative pressure, sucking suspected accumulated water or high-humidity gas inside the air velocity meter to an external water collection container or drying device. It can be understood that the opening degree adjustment of the miniature electronically controlled valve is based on a preset mapping relationship, which is expressed by the formula:

[0104]

[0105] in: This indicates the target opening degree of the miniature electronically controlled valve. This indicates the suggested water absorption flow rate obtained from the analysis. The function represents the real-time negative pressure value generated by the operation of the miniature negative pressure pump. Characterized the ability to maintain the recommended water intake flow rate Miniature electronically controlled valve opening With real-time negative pressure value The system dynamically adjusts the relationship between these parameters. During active water absorption, the system continuously acquires updated cavity humidity and airflow pressure difference values. It analyzes the rate of decrease in cavity humidity and the trend of the airflow pressure difference returning to its standard value to assess the effectiveness of the water absorption process in real time. In some embodiments, if the rate of decrease in cavity humidity is lower than the expected threshold or the airflow pressure difference does not show a trend of returning to its standard value within a preset time, the system generates a record and associates this situation with the effectiveness evaluation log of the current water absorption strategy.

[0106] See Figure 4 This is a multi-parameter comparison bar chart showcasing the active water intake strategy of a UAV's airspeed indicator. It visually presents the differences in recommended water intake velocity, electronic valve opening, and negative pressure values ​​for different target water intake areas, reflecting the system's refined water intake control logic for different regions. Areas directly in contact with the external airflow, such as the airspeed tube inlet and the cavity sidewalls, have higher configured water intake parameters, consistent with their susceptibility to external moisture and high condensation risk. This chart clearly demonstrates the system's regionalized and precise water intake strategy, its value lying in directly correlating water intake intensity with the risk of regional water accumulation, avoiding resource waste. Through coordinated parameter adjustment, it ensures effective water intake while avoiding interference with the airspeed indicator's aerodynamic measurements. Transforming the abstract water intake strategy into an intuitive parameter comparison facilitates understanding and verification of the system logic by R&D and maintenance personnel.

[0107] In one embodiment of the present invention, after the active water absorption program is initiated, if the updated cavity humidity value is still higher than the preset safe humidity threshold, or the airflow pressure difference value has not fully returned to the normal range, it is determined that auxiliary drying is required and an auxiliary drying program initiation command is generated. This command includes a suggested drying temperature and drying duration. The system activates the electrothermal film or micro-electrothermal element built into or near the airspeed sensor cavity wall, and controls the electrothermal film or micro-electrothermal element to heat according to the suggested drying temperature to assist in drying the residual moisture inside the airspeed sensor. At the same time, the micro negative pressure pump is kept running at low power to continuously discharge the heated and evaporated water vapor. During the auxiliary drying process, the internal temperature reading and the updated cavity humidity value are continuously monitored, and the drying parameters are dynamically adjusted according to the humidity decrease. When a maintenance alarm command is generated, the system immediately interrupts any ongoing water absorption or drying processes, assembles an alarm message, and includes at least the real-time probability of water accumulation inside the airspeed indicator, an estimated water distribution area, an estimated water volume, the last valid humidity status characteristics, and abnormal airflow characteristics. This alarm message is then sent to the ground control station or a handheld terminal for maintenance personnel via the UAV's onboard data link or a dedicated maintenance interface. All relevant data for this alarm event, including timestamps, all sensor data sequences, and diagnostic model outputs, is recorded in the system's local storage. Finally, the system switches to safety monitoring mode.

[0108] In specific implementation, after the active water absorption program is initiated, the system continuously monitors and updates the cavity humidity value and airflow pressure difference value. If the updated cavity humidity value is still higher than the preset safe humidity threshold or the airflow pressure difference value has not fully returned to the normal range, the system determines that auxiliary drying is required. The system generates an instruction to initiate the auxiliary drying program, which includes a suggested drying temperature and drying duration. In some embodiments, the system determines the specific values ​​of the suggested drying temperature and drying duration based on the difference between the last obtained cavity humidity value and the safe humidity threshold, as well as the degree to which the airflow pressure difference value deviates from the normal range, using a built-in lookup table method. The mapping table used in the lookup table method is pre-calibrated based on the thermal properties of the air velocity meter material and experimental data on drying efficiency.

[0109] In practice, the system activates an electrothermal film or miniature electrothermal element embedded in or near the airspeed sensor cavity wall. The system controls this film or element to heat the airspeed sensor to the recommended drying temperature as instructed in the auxiliary drying program, thus assisting in drying residual moisture inside the airspeed sensor. Simultaneously, the system maintains a miniature negative pressure pump operating at low power to continuously expel the heated and evaporated water vapor. This low-power operation means that the pump's speed or suction force is set to a fixed value lower than during the active water suction program phase. This fixed value is sufficient to maintain slow gas flow and remove water vapor without significantly interfering with the airspeed sensor's normal aerodynamic measurements. During the auxiliary drying process, the system continuously monitors the internal temperature readings and updated cavity humidity values, dynamically adjusting the drying parameters based on the humidity decrease.

[0110] In practical implementation, when a maintenance alarm command is generated, the system immediately interrupts any ongoing water absorption or drying processes. The system assembles an alarm information message, which includes at least the real-time probability of water accumulation inside the airspeed indicator, the estimated water distribution area, the estimated water volume, and the last valid humidity status characteristics and airflow anomaly characteristics. Optionally, the last valid humidity status characteristics and airflow anomaly characteristics refer to the average humidity level, humidity gradient trend, condensation risk index, pressure difference fluctuation amplitude, airflow stability coefficient, and blockage tendency assessment value output by the feature analysis module in the most recent complete analysis cycle before the maintenance alarm command is generated. The system sends the assembled alarm information message to the ground control station or the handheld terminal of the maintenance personnel via the UAV's onboard data link or a dedicated maintenance interface. In practical implementation, the system records all relevant data for this alarm event in its local storage unit, including timestamps, all sensor data sequences, and diagnostic model output results. The system switches its status to a safe monitoring mode, in which it only maintains basic environmental monitoring functions and awaits external maintenance intervention commands. It is understandable that the basic environmental monitoring function refers to continuing to collect and store cavity humidity values, airflow pressure difference values, and internal temperature readings at a lower frequency, but no longer performing a complete cycle of feature analysis, water accumulation diagnosis, and generating any disposal instructions.

[0111] See Figure 5This is a bar chart analyzing the alarm triggering conditions of a drone airspeed indicator water accumulation detection system. By comparing the probability and volume of water accumulation under different operating scenarios, and combining this with two alarm threshold lines, it clearly demonstrates the judgment logic for alarm triggering. This chart intuitively illustrates the system's hierarchical alarm logic, with dual threshold judgments using both water accumulation probability and volume as the basis for judgment, avoiding misjudgments based on a single indicator. Different scenarios correspond to different handling strategies, improving the system's reliability and safety. It can directly verify whether the output of the water accumulation diagnostic model meets expectations, verifying the model's accuracy and robustness by comparing the matching degree between water accumulation probability, volume, and alarm thresholds under different scenarios. Figure 5 The system clearly distinguishes the risk levels of different scenarios, ensuring that high-risk scenarios such as severe water accumulation and sensor malfunctions can be identified and alarms triggered in a timely manner, while scenarios such as moderate water accumulation will enter an active water absorption process, avoiding excessive alarms or missed alarms and ensuring the flight safety of drones.

[0112] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A smart water accumulation detection and absorption system for the internal water speedometer of a drone, characterized in that, The system includes: The data acquisition module acquires multi-source environmental sensing data from inside the UAV airspeed meter. The multi-source environmental sensing data includes the humidity value of the airspeed tube cavity collected by the humidity sensor, the air pressure difference value collected by the micro-pressure differential sensor, and the internal temperature reading collected by the temperature sensor. The feature analysis module performs fusion preprocessing analysis on the multi-source environmental sensing data to generate humidity state characteristics and airflow anomaly characteristics inside the airspeed meter. The humidity state characteristics include average humidity level, humidity gradient change trend and condensation risk index. The airflow anomaly characteristics include pressure difference fluctuation amplitude, airflow stability coefficient and blockage tendency assessment value. The water accumulation diagnosis module inputs the humidity state characteristics and airflow anomaly characteristics into the pre-trained water accumulation diagnosis model to calculate the real-time water accumulation probability inside the airspeed indicator, the estimated value of the water accumulation distribution area, and the estimated water accumulation volume. The treatment control module generates graded treatment instructions based on the real-time water accumulation probability inside the air velocity meter, the estimated value of the water accumulation distribution area, and the estimated water volume. The graded treatment instructions include instructions to start the active water absorption program, instructions to start the auxiliary drying program, or instructions to trigger a maintenance alarm. The process of fusing and preprocessing the multi-source environmental sensor data to generate humidity state characteristics and airflow anomaly characteristics within the airspeed sensor includes: The cavity humidity value is subjected to time series smoothing filtering to eliminate instantaneous measurement noise and obtain a smoothed cavity humidity sequence; Calculate the average value of the smoothed cavity humidity sequence over a preset analysis period, and use it as the average humidity level; The smoothed cavity humidity sequence is analyzed for its rate of change and second derivative during the analysis period to determine whether the humidity is continuously rising, falling, or fluctuating. A humidity change curve is then fitted, and the humidity gradient change trend is extracted from it. By combining the internal temperature readings with the smoothed cavity humidity sequence, and based on the dew point temperature calculation model, the probability of condensation occurring on the internal surface of the airspeed meter under the current internal temperature and humidity conditions is assessed, and the condensation risk index is calculated. Analyze the fluctuation of the airflow pressure difference over time, calculate its standard deviation and peak-to-peak value, and obtain the amplitude of the pressure difference fluctuation. The airflow stability coefficient is calculated based on the degree of deviation between the airflow pressure difference value and the standard pressure difference value under the theoretical no-water-accumulation condition, as well as the frequency characteristics of the pressure difference change. Based on the variation pattern of the airflow pressure difference value, combined with the fluid dynamics model, it is determined whether there is a local decrease in flow velocity or airflow separation phenomenon in the airspeed tube, and the blockage tendency assessment value characterizing local blockage is generated.

2. The intelligent detection and absorption system for water accumulation inside the UAV airspeed meter according to claim 1, characterized in that, The method for calculating the condensation risk index includes: Based on the internal temperature reading and the smoothed cavity humidity sequence, the dew point temperature under the current condition is calculated using the internationally recognized dew point calculation formula. The difference between the internal temperature reading and the dew point temperature is calculated to obtain the superheat. Establish a mapping table between superheat and condensation risk index. In the mapping table, the smaller the superheat, the higher the condensation risk index. Based on the calculated superheat value, the corresponding condensation risk index is obtained by querying the mapping table. Meanwhile, the condensation risk index is spatially weighted and corrected by taking into account the local temperature differences in different parts of the airspeed gauge to more accurately reflect the risk level in different areas.

3. The intelligent detection and absorption system for water accumulation inside the UAV airspeed meter according to claim 2, characterized in that, The step of inputting the humidity state characteristics and airflow anomaly characteristics into a pre-trained water accumulation diagnosis model to calculate the real-time water accumulation probability inside the airspeed indicator, the estimated water accumulation area, and the estimated water accumulation volume includes: The water accumulation diagnostic model receives the input of the average humidity level, humidity gradient change trend, condensation risk index, pressure difference fluctuation amplitude, airflow stability coefficient, and blockage tendency assessment value. The water accumulation diagnosis model, based on its internal multi-layer neural network structure, first performs feature fusion and dimensionality reduction to generate a comprehensive state vector. The comprehensive state vector is matched with the state vectors in the historical water accumulation case library for similarity, and the real-time water accumulation probability inside the airspeed meter is calculated using the built-in logistic regression classifier of the model. The probability value represents the possibility that water accumulation exists at present. Meanwhile, the water accumulation diagnostic model estimates the location of water accumulation by using an interpolation algorithm based on the spatial correlation between the condensation risk index and the blockage tendency assessment value, combined with the three-dimensional model of the internal structure of the air velocity meter, and outputs the estimated value of the water accumulation distribution area. The water accumulation diagnostic model further calculates the estimated water accumulation volume by using the fluid volume function in conjunction with the deviation between the airflow pressure difference value and the theoretical dry pressure difference value, combined with the internal geometric parameters of the airspeed tube.

4. The intelligent detection and absorption system for water accumulation inside the UAV airspeed meter according to claim 3, characterized in that, The water accumulation diagnostic model, based on the spatial correlation between the condensation risk index and the blockage tendency assessment value, and combined with a three-dimensional model of the internal structure of the airspeed indicator, estimates the location of water accumulation using an interpolation algorithm, and outputs an estimated value of the water accumulation distribution area, specifically including: Obtain the coordinate information of each position in the three-dimensional model of the internal structure of the airspeed indicator; The condensation risk index is mapped to the corresponding sensor installation location in the three-dimensional model of the internal structure of the airspeed meter; The blockage tendency assessment value is mapped to the location of the corresponding airflow monitoring point in the three-dimensional model of the internal structure of the airspeed indicator; Calculate the spatial correlation weight between the condensation risk index and the blockage tendency assessment value, wherein the correlation weight is determined based on the distance and numerical correlation between the two feature values ​​in three-dimensional space; Based on the associated weights, bilinear interpolation calculations are performed on the mesh nodes of the three-dimensional model of the internal structure of the airspeed indicator. A water accumulation probability distribution map of each grid node inside the airspeed meter is generated using an interpolation algorithm; Extract continuous regions with a water accumulation probability higher than a set threshold from the water accumulation probability distribution map, and use them as estimates of the water accumulation distribution areas.

5. The intelligent detection and absorption system for water accumulation inside the UAV airspeed meter according to claim 1, characterized in that, The training and updating methods for the pre-trained water accumulation diagnosis model include: In a laboratory or controlled environment, different degrees of internal water accumulation or high humidity were artificially simulated for different types of UAV airspeed meter samples. When simulating various states, corresponding multi-source environmental sensor data are collected synchronously, including cavity humidity value, airflow pressure difference value and internal temperature reading, and the corresponding real water accumulation situation is recorded as a label. The real water accumulation situation includes whether there is water accumulation, the approximate area and volume range of water accumulation. The collected labeled sensor data is used to conduct supervised training on the initial neural network model. The model parameters are adjusted through the backpropagation algorithm until the error between the model output water accumulation probability, area estimation and volume estimation and the real label meets the accuracy requirements. Deploy the trained model to a real drone system; In actual operation, when the system executes the aforementioned maintenance alarm command and the maintenance personnel conduct on-site handling and confirm the water accumulation, the maintenance confirmation result is used as a new real label, which together with the sensor data recorded at that time constitutes a new training sample. After periodically or accumulating a certain number of new samples, the water accumulation diagnosis model is incrementally learned or fine-tuned to continuously optimize model performance.

6. The intelligent detection and absorption system for water accumulation inside the UAV airspeed meter according to claim 1, characterized in that, The step of generating tiered treatment instructions based on the real-time probability of water accumulation inside the airspeed sensor, the estimated value of the water accumulation distribution area, and the estimated water volume includes: The probability of water accumulation inside the airspeed meter in real time is compared with a first probability threshold and a second probability threshold, wherein the first probability threshold is lower than the second probability threshold. When the probability of water accumulation inside the airspeed meter is lower than the first probability threshold, it is determined that there is no water accumulation or only a trace of moisture, a monitoring command is generated, and the system maintains its normal monitoring state. When the real-time probability of water accumulation inside the airspeed meter is between the first probability threshold and the second probability threshold, it is determined that there is an initial risk of water accumulation or moderate humidity. In this case, if the estimated water volume is less than a preset volume threshold, the command to start the active water absorption program is generated. If the estimated volume of water accumulation is greater than or equal to the preset volume threshold, it is determined that there is a significant risk of water accumulation, and the aforementioned maintenance alarm command is generated. When the probability of water accumulation inside the airspeed meter is higher than the second probability threshold, it is directly determined that there is a high probability of water accumulation or a serious risk, and the maintenance alarm command is generated. For cases where an initial risk of water accumulation or moderate humidity is identified and the estimated water volume is less than the volume threshold, a water absorption strategy for a specific area is embedded in the active water absorption program command, based on the estimated water distribution area.

7. The intelligent detection and absorption system for water accumulation inside the UAV airspeed meter according to claim 6, characterized in that, After generating the instruction to initiate the active water absorption process, the system further includes executing the active water absorption process: The active water absorption program command is analyzed to determine the water absorption strategy for a specific area, and the target water absorption area and the recommended water absorption flow rate are determined. Activate the micro electronically controlled valve and the micro negative pressure pump located near the target water absorption area, wherein the micro negative pressure pump is connected to a specific suction port of the internal cavity of the air velocity meter through a pipeline; The opening degree of the micro electronically controlled valve is controlled to match the recommended water suction flow rate, and the micro negative pressure pump is started to generate negative pressure to draw the suspected water or high humidity gas inside the air velocity meter to the external water collection container or drying device. During the active water absorption process, the humidity value of the cavity and the airflow pressure difference value are continuously and synchronously acquired and updated. The effectiveness of the water absorption treatment is evaluated in real time by analyzing the rate of decrease in the humidity value of the cavity and the trend of the airflow pressure difference value recovering to the standard value.

8. The intelligent detection and absorption system for water accumulation inside the UAV airspeed meter according to claim 7, characterized in that, The execution of the instruction to start the auxiliary drying program includes: If, after the active water absorption process is executed, the updated humidity value of the cavity is still higher than the preset safe humidity threshold, or the airflow pressure difference value has not fully returned to the normal range, then it is determined that auxiliary drying is required. Generate the instruction to start the auxiliary drying program, which includes a suggested drying temperature and drying duration; Activate the electrothermal film or miniature electrothermal element built into or near the airspeed gauge cavity wall; The electric heating film or micro electric heating element is controlled to heat at the recommended drying temperature to assist in drying the residual moisture inside the airspeed meter. At the same time, the miniature negative pressure pump is kept running at low power to continuously discharge the water vapor that has been heated and evaporated; During the assisted drying process, the internal temperature readings and the updated humidity values ​​of the cavity are continuously monitored to ensure that the temperature does not exceed the upper limit of the air velocity meter material's tolerance, and the drying parameters are dynamically adjusted according to the decrease in humidity.

9. The intelligent detection and absorption system for water accumulation inside the UAV airspeed meter according to claim 8, characterized in that, The execution of the trigger maintenance alarm command includes: When the aforementioned maintenance alarm command is generated, the system immediately interrupts any ongoing water absorption or drying process; Assemble an alarm information message, the alarm information message including at least the real-time probability of water accumulation inside the airspeed meter, the estimated value of the water accumulation distribution area, the estimated amount of water accumulation volume, the last valid humidity status characteristics, and the airflow anomaly characteristics; The alarm information message is sent to the ground control station or the handheld terminal of the maintenance personnel via the airborne data link of the UAV or a dedicated maintenance interface; Record all relevant data for this alarm event in the system's local storage unit, including timestamps, all sensor data sequences, and diagnostic model output results; Switch the system status to security monitoring mode. In this mode, only basic environmental monitoring functions are maintained, and the system awaits external maintenance intervention instructions.

Images